
import pandas as pd
import numpy as np

df = pd.read_csv('airbnb-listings.csv', delimiter=";")


df.head()


df.columns.values

array(['ID', 'Listing Url', 'Scrape ID', 'Last Scraped', 'Name',
       'Summary', 'Space', 'Description', 'Experiences Offered',
       'Neighborhood Overview', 'Notes', 'Transit', 'Access',
       'Interaction', 'House Rules', 'Thumbnail Url', 'Medium Url',
       'Picture Url', 'XL Picture Url', 'Host ID', 'Host URL',
       'Host Name', 'Host Since', 'Host Location', 'Host About',
       'Host Response Time', 'Host Response Rate', 'Host Acceptance Rate',
       'Host Thumbnail Url', 'Host Picture Url', 'Host Neighbourhood',
       'Host Listings Count', 'Host Total Listings Count',
       'Host Verifications', 'Street', 'Neighbourhood',
       'Neighbourhood Cleansed', 'Neighbourhood Group Cleansed', 'City',
       'State', 'Zipcode', 'Market', 'Smart Location', 'Country Code',
       'Country', 'Latitude', 'Longitude', 'Property Type', 'Room Type',
       'Accommodates', 'Bathrooms', 'Bedrooms', 'Beds', 'Bed Type',
       'Amenities', 'Square Feet', 'Price', 'Weekly Price',
       'Monthly Price', 'Security Deposit', 'Cleaning Fee',
       'Guests Included', 'Extra People', 'Minimum Nights',
       'Maximum Nights', 'Calendar Updated', 'Has Availability',
       'Availability 30', 'Availability 60', 'Availability 90',
       'Availability 365', 'Calendar last Scraped', 'Number of Reviews',
       'First Review', 'Last Review', 'Review Scores Rating',
       'Review Scores Accuracy', 'Review Scores Cleanliness',
       'Review Scores Checkin', 'Review Scores Communication',
       'Review Scores Location', 'Review Scores Value', 'License',
       'Jurisdiction Names', 'Cancellation Policy',
       'Calculated host listings count', 'Reviews per Month',
       'Geolocation', 'Features'], dtype=object)


len(df.axes[1])

89


len(df.axes[0])

494954


df = df.filter(['Longitude','Latitude','Space','Accommodates','City','Host Response Time','Host Verifications','Street','Property Type','Room Type','Bathrooms','Bedrooms','Beds','Price','Cleaning Fee','Review Scores Rating','Cancellation Policy'])


df =df.dropna()


#ON CHECK LES NOMS DES VILLES
df['City'].value_counts()

Paris                                  17954
London                                 15581
Los Angeles                             9008
Barcelona                               7316
New York                                7056
                                       ...  
Royal Arsenal Riverside                    1
Streatham Hill, London, England, GB        1
Palmers Green                              1
Leyton                                     1
Афины                                      1
Name: City, Length: 3087, dtype: int64


#ON GARDE UNIQUEMENT PARIS
df.drop( df[ df['City'] != "Paris" ].index, inplace=True)


df['Property Type'].value_counts()

Apartment             17249
Loft                    271
House                   197
Bed & Breakfast          80
Condominium              42
Other                    35
Townhouse                25
Guesthouse               22
Boat                     21
Boutique hotel            2
Serviced apartment        2
Earth House               2
Treehouse                 1
Timeshare                 1
Hostel                    1
Tipi                      1
Camper/RV                 1
Dorm                      1
Name: Property Type, dtype: int64


df.drop( df[ df['Property Type'] != "Apartment" ].index, inplace=True)


df['Street'].value_counts()

Montmartre, Paris, Île-de-France 75018, France                        956
Saint-Germain-des-Prés - Odéon, Paris, Île-de-France 75006, France    614
Batignolles, Paris, Île-de-France 75017, France                       485
Bastille, Paris, Île-de-France 75011, France                          460
République, Paris, Île-de-France 75011, France                        457
                                                                     ... 
Paris, Île de France 75010, France                                      1
Gare du Nord - Gare de I'Est, Paris, Ile-de-France 75010, France        1
Paris, Ãle-de-France 75019, France                                     1
Paris, Ile de France 75019, France                                      1
Champs-Elysées, Paris, Ile-de-France 75008, France                      1
Name: Street, Length: 413, dtype: int64


df['Street']=df['Street'].str.extract('(\d+)')
df =df.dropna()
df['Street'] = df['Street'].astype(int)
df.drop( df[ (df['Street']< 75001) | (df['Street']> 75020) ].index, inplace=True)
df['Street'] = df['Street'] - 75000


df['Street'].value_counts()

18    1739
11    1588
10    1201
15    1111
3      994
17     977
4      900
6      822
5      797
9      771
2      742
20     711
7      656
19     627
14     567
8      525
1      513
12     511
13     428
16     411
Name: Street, dtype: int64


df['Host Response Time'].value_counts()

within an hour        8207
within a few hours    4259
within a day          3656
a few days or more     469
Name: Host Response Time, dtype: int64


df['Room Type'].value_counts()

Entire home/apt    15406
Private room        1115
Shared room           70
Name: Room Type, dtype: int64


df['Cancellation Policy'].value_counts()

strict             8511
moderate           5586
flexible           2492
super_strict_30       2
Name: Cancellation Policy, dtype: int64


response_type = {
    'within an hour' : 1,
    'within a few hours' : 2,
    'within a day' : 3,
    'a few days or more' : 4,
}

nw_list_response = []

for i in df['Host Response Time']:
    nw_list_response.append(response_type[i])

df['Host Response Time'] = nw_list_response


room_type = {
    'Entire home/apt' : 1,
    'Private room' : 2,
    'Shared room' : 3,
}

nw_list_room = []

for i in df['Room Type']:
    nw_list_room.append(room_type[i])

df['Room Type'] = nw_list_room


cancel_policy = {
    'strict' : 1,
    'flexible' : 2,
    'moderate' : 3,
    'super_strict_30' : 4,
}

nw_list_cancelpolicy= []

for i in df['Cancellation Policy']:
    nw_list_cancelpolicy.append(cancel_policy[i])

df['Cancellation Policy'] = nw_list_cancelpolicy


df['Space'].iloc[2]

'This charming 25m2 studio is very bright and colourful. With a bohemian feel, has everything for 2 guests to feel comfortable and cozy.  On a 1st floor without elevator, on a great part of Paris, easy access to Metro Stations and bus stops at the doorstep. Supermarket  and bakery just across the street and fabulous restaurants all around in less than 2 minutes walk.'


df=df[df["Space"].str.contains("m2")==True]
df['Surface']=df.Space.str.split("m2", expand=True)[0]
df['Surface']=df['Surface'].str.split().str[-1]
df['Surface']=df['Surface'].str.extract('(\d+)')


df['Host Verifications'].value_counts()

email,phone,reviews,jumio                                                            879
email,phone,reviews                                                                  671
email,phone,facebook,reviews,jumio                                                   219
email,phone,reviews,jumio,government_id                                              178
email,phone,facebook,reviews                                                         141
                                                                                    ... 
email,phone,facebook,amex,reviews,kba,work_email                                       1
email,phone,linkedin,reviews,jumio,offline_government_id,government_id,work_email      1
email,phone,facebook,google,reviews,jumio                                              1
email,phone,google,reviews,jumio,offline_government_id,government_id,work_email        1
email,phone,facebook,linkedin,reviews,jumio,work_email                                 1
Name: Host Verifications, Length: 80, dtype: int64


df['Host Verifications']= df['Host Verifications'].str.replace(',', ' ')
df['Host Verifications'] = df['Host Verifications'].apply(lambda n: len(n.split()))


df['Host Verifications'].value_counts()

4    1066
3     681
5     590
6     238
7      73
2      28
8      12
9       4
1       3
Name: Host Verifications, dtype: int64


df.head()


len(df)

2695


df.dtypes

Longitude               float64
Latitude                float64
Space                    object
Accommodates            float64
City                     object
Host Response Time        int64
Host Verifications        int64
Street                    int32
Property Type            object
Room Type                 int64
Bathrooms               float64
Bedrooms                float64
Beds                    float64
Price                   float64
Cleaning Fee            float64
Review Scores Rating    float64
Cancellation Policy       int64
Surface                  object
dtype: object


df = df.drop(columns=['City', 'Property Type','Space'], axis=1)


df = df.rename({'Host Verifications': 'Host Credibility Degree', 'Street': 'District','Review Scores Rating':'Review'}, axis=1)


df =df.dropna()
df=df.reset_index()
df = df.drop(columns=['index'], axis=1)

df


import seaborn as sns
import matplotlib.pyplot as plt


df.plot(kind="scatter", 
                  x="Longitude", y="Latitude", 
                  alpha=0.4, figsize=(14,7), 
                  c="Price", cmap="gist_heat_r", 
                  colorbar=True, sharex=False);


import plotly.express as px
import pandas as pd

#df = pd.read_csv("location_coordinate.csv")

fig = px.scatter_geo(df,lat='Latitude',lon='Longitude', hover_name="Price",color='Price',scope='europe',height=800, width=1000,center=dict(lat=48.855, lon=2.325))
fig.update_layout(title = 'World map', title_x=0.5)

fig.show()


df.boxplot(column =['Price'], grid = False)

df['Price'].describe()

count    2681.000000
mean      100.458038
std        66.847296
min        20.000000
25%        60.000000
50%        80.000000
75%       120.000000
max       800.000000
Name: Price, dtype: float64


# Calculate the IQR
q1 = df['Price'].quantile(0.25)
q3 = df['Price'].quantile(0.75)
iqr = q3 - q1

# Calculate the lower and upper boundaries
lower_bound = q1 - (1.5 * iqr)
upper_bound = q3 + (1.5 * iqr)

# Filter the DataFrame to only include rows where 'col' is within the boundaries
df = df[(df['Price'] > lower_bound) & (df['Price'] < upper_bound)]


sns.barplot(x="Room Type", y="Price", data=df)
plt.title('Price according to the type of room')
plt.show()


sns.barplot(x="District", y="Price", data=df)
plt.title('Price according to the district')
plt.show()


plt.scatter(x="Review", y="Price", data=df)
plt.title('Price according to the rating')
plt.show()


sns.barplot(x="Host Credibility Degree", y="Price", data=df)
plt.title('Price according to the degree of host verirification')
plt.show()


sns.barplot(x="Host Response Time", y="Review", data=df)
plt.title('Price according to the Host Response Time')
plt.show()


df_ml = df.copy()


from sklearn.linear_model import LinearRegression
from sklearn.linear_model import Lasso
from sklearn.linear_model import Ridge
from sklearn.linear_model import ElasticNet
from sklearn.linear_model import BayesianRidge
from sklearn.linear_model import ARDRegression
from sklearn.linear_model import TweedieRegressor
from sklearn.linear_model import OrthogonalMatchingPursuit
from sklearn.linear_model import TheilSenRegressor
from sklearn.linear_model import HuberRegressor
from sklearn.ensemble import RandomForestRegressor
from sklearn.ensemble import GradientBoostingRegressor
from sklearn.ensemble import AdaBoostRegressor
from sklearn.ensemble import BaggingRegressor

from sklearn import linear_model
from sklearn.metrics import mean_squared_error
from sklearn.model_selection import train_test_split


models = [LinearRegression(), Lasso(alpha=0.1), Ridge(alpha=1.2), ElasticNet(alpha=0.01), BayesianRidge(), ARDRegression(), OrthogonalMatchingPursuit(), TweedieRegressor(max_iter=300), TheilSenRegressor(), HuberRegressor(max_iter=3000), RandomForestRegressor(max_depth=8), linear_model.Lars(n_nonzero_coefs=7, normalize=False), GradientBoostingRegressor(random_state=3), AdaBoostRegressor(random_state=0, n_estimators=100), BaggingRegressor()]


models_name = ['LinearRegression', 'Lasso', 'Ridge', 'ElasticNet', 'BayesianRidge', 'ARDRegression', 'OrthogonalMatchingPursuit', 'TweedieRegressor', 'TheilSenRegressor', 'HuberRegressor', 'RandomForestRegressor', 'LARS', 'GradientBoostingRegressor', 'AdaBoostRegressor', 'BaggingRegressor']


values_r2 = [] #the list that will store the r-squared values
values_root = [] #the list that will store the rmse values


for model in models: #we iterate for each models in the list
    
    y = df_ml['Price'].values #y is the feature we want to predict
    x = df_ml.drop('Price', 1).values #x is the dataset that will predict the y

    X_train, X_test, y_train, y_test = train_test_split(x, y, test_size = 0.30, random_state=42) #we split the dataset

    reg = model
    reg.fit(X_train, y_train) #we fit the dataset

    y_pred = reg.predict(X_test)

    values_r2.append(reg.score(X_test, y_test)) #we put the r-squared into the r2 list
    rmse = np.sqrt(mean_squared_error(y_test, y_pred)) #we calculate the rmse
    values_root.append(rmse) #and then put it into the rmse list


#we create dataframes with the lists
r2 = pd.DataFrame(values_r2, columns=['r2']) 
r2['model'] = models_name

rt = pd.DataFrame(values_root, columns=['RootMeanSquaredError'])
rt['model'] = models_name


sns.barplot(x="r2", y="model",data=r2, order=r2.sort_values(['r2']).model)
plt.xlim(0,0.95)
plt.title('r-squared value according to the regression model used')
plt.show()


sns.barplot(x="RootMeanSquaredError", y="model",data=rt, order=rt.sort_values(['RootMeanSquaredError']).model.iloc[::-1])
plt.xlim(0.1,100)
plt.title('rmse value according to the regression model used')
plt.show()


y = df_ml['Price'].values
x = df_ml.drop('Price', 1).values

X_train, X_test, y_train, y_test = train_test_split(x, y, test_size = 0.3, random_state=42)

reg = GradientBoostingRegressor()
reg.fit(X_train, y_train)

pred = reg.predict(x)

#the error is the difference between the value calculated by the prediction and the acutal value
error = (abs(df_ml['Price'] - pred))/df_ml['Price']*100


error_mean = sum(error)/len(error)

error_mean

6.671104921196848

	ID	Listing Url	Scrape ID	Last Scraped	Name	Summary	Space	Description	Experiences Offered	Neighborhood Overview	...	Review Scores Communication	Review Scores Location	Review Scores Value	License	Jurisdiction Names	Cancellation Policy	Calculated host listings count	Reviews per Month	Geolocation	Features
0	4008728	https://www.airbnb.com/rooms/4008728	20170402075052	2017-04-02	Luxurious 3 bedroom, centrum, 180m2	‪This luxurious apartment is situated in the c...	In the heart of hip & lively Amsterdam Centrum...	‪This luxurious apartment is situated in the c...	none	There is a wide variety of cafes and restauran...	...	10.0	10.0	9.0	NaN	Amsterdam	strict	1.0	1.52	52.36523660409065,4.878249575101092	Host Has Profile Pic,Host Identity Verified,Is...
1	7778612	https://www.airbnb.com/rooms/7778612	20170402075052	2017-04-02	Luxury apartment in city centre	Beautiful, large (105m2, 2 floors) and quiet a...	1 living room 1 dining room 2 bedrooms 1 bathr...	Beautiful, large (105m2, 2 floors) and quiet a...	none	Within a few minutes walking you'll be in the ...	...	10.0	10.0	9.0	NaN	Amsterdam	strict	1.0	0.81	52.367309030897516,4.8738407410699915	Host Is Superhost,Host Has Profile Pic,Host Id...
2	8264596	https://www.airbnb.com/rooms/8264596	20170402075052	2017-04-02	Cosy apartment across Vondelpark	This central located apartment will make you f...	NaN	This central located apartment will make you f...	none	NaN	...	10.0	10.0	10.0	NaN	Amsterdam	flexible	1.0	0.05	52.361943953804776,4.866686522141489	Host Has Profile Pic,Host Identity Verified,Is...
3	2180729	https://www.airbnb.com/rooms/2180729	20170402075052	2017-04-02	Spacious City Apartment Oud-West	Living like an Amsterdam resident in this ligh...	There are 2 bedrooms each with a double bed. I...	Living like an Amsterdam resident in this ligh...	none	The 85 m2 apartment is located on a quiet stre...	...	9.0	9.0	9.0	NaN	Amsterdam	flexible	1.0	0.62	52.37014616115532,4.866281767968382	Host Has Profile Pic,Host Identity Verified,Is...
4	14463171	https://www.airbnb.com/rooms/14463171	20170402075052	2017-04-02	Cosy Studio Apartment Center Amsterdam	Studio Apartment Centre Amsterdam Perfect stud...	NaN	Studio Apartment Centre Amsterdam Perfect stud...	none	NaN	...	8.0	9.0	9.0	NaN	Amsterdam	moderate	1.0	2.04	52.368817829014425,4.871249476925455	Host Has Profile Pic

	Longitude	Latitude	Accommodates	Host Response Time	Host Credibility Degree	District	Room Type	Bathrooms	Bedrooms	Beds	Price	Cleaning Fee	Review	Cancellation Policy	Surface
0	2.365122	48.855027	2.0	2	4	4	1	1.0	0.0	1.0	80.0	40.0	94.0	1	25
1	2.357037	48.857787	4.0	3	4	4	1	1.0	2.0	2.0	94.0	26.0	80.0	1	28
2	2.353065	48.858664	4.0	3	4	4	1	1.0	1.0	1.0	134.0	50.0	98.0	1	40
3	2.355115	48.852292	2.0	1	3	4	1	1.0	1.0	1.0	135.0	45.0	96.0	1	32
4	2.353662	48.890418	2.0	4	5	18	1	1.0	1.0	1.0	60.0	20.0	91.0	3	45
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
2676	2.387538	48.848758	8.0	1	4	12	1	2.0	4.0	6.0	229.0	70.0	97.0	1	150
2677	2.389579	48.840585	2.0	3	5	12	1	1.0	1.0	1.0	45.0	30.0	96.0	3	27
2678	2.302691	48.880195	8.0	1	4	17	1	2.0	3.0	4.0	160.0	90.0	78.0	1	85
2679	2.306053	48.857639	4.0	1	6	7	1	1.0	1.0	2.0	155.0	35.0	97.0	1	40
2680	2.326907	48.851534	4.0	3	4	6	1	1.0	1.0	3.0	136.0	117.0	100.0	1	45

Appartments in Paris¶

Import data and useful librairies¶

The dataset¶

At this step we have a DataFrame with 89 columns and 494 954 raws¶

Cleaning the dataset¶

Now we will explore 3 different columns by displaying their values¶

The Space column is a little bit technical to clean¶

The Host Verifications column¶

Visualizing the data¶

First of all, we need to explore the Price to see if there is any outlier¶

Prediction of price¶

Visualizing the models¶

GradientRegressor¶

Conclusion¶

Limits and critics¶

	Longitude	Latitude	Space	Accommodates	City	Host Response Time	Host Verifications	Street	Property Type	Room Type	Bathrooms	Bedrooms	Beds	Price	Cleaning Fee	Review Scores Rating	Cancellation Policy	Surface
24345	2.365122	48.855027	This charming 25m2 studio is very bright and c...	2.0	Paris	2	4	4	Apartment	1	1.0	0.0	1.0	80.0	40.0	94.0	1	25
24352	2.357037	48.857787	28m2, 2 Separated Mezzanines, one kitchen, one...	4.0	Paris	3	4	4	Apartment	1	1.0	2.0	2.0	94.0	26.0	80.0	1	28
24387	2.353065	48.858664	A 430 sq ft / 40 m2, one bedroom apartment is ...	4.0	Paris	3	4	4	Apartment	1	1.0	1.0	1.0	134.0	50.0	98.0	1	40
24397	2.355115	48.852292	Hello future friends, I would like to describe...	2.0	Paris	1	3	4	Apartment	1	1.0	1.0	1.0	135.0	45.0	96.0	1	32
24563	2.353662	48.890418	Duplex full of light in the middle of the Gout...	2.0	Paris	4	5	18	Apartment	1	1.0	1.0	1.0	60.0	20.0	91.0	3	45